Tableau on Tour - Robert Kosara – The Plot Thickens: Using Visualisation to Tell Stories about Data

Whilst judging a Visualisation – a fellow judge described something as ‘Just Visualisation’ and this started Robert thinking about what the levels of understanding include:

1.      Visualisation

2.      Context

3.      Story

New York Times wouldn’t use scatterplots 2 years ago but they are now used routinely as they were found to be understood.

Scatterplots – humans are good at drawing a trendline through a scatterplot but can’t draw the true diagonal line

-        Adding graduated lines to show trend lines at x% change helps the consumer appreciate the trend they are seeing (example from Hannah Fairfield and Graham Roberts at the New York Times - http://www.nytimes.com/interactive/2009/03/01/business/20090301_WageGap.html?_r=0)

-        Allows the reader to conduct analysis that was previously impossible

The ‘stepper’ in Tableau Storypoints is taken from News Media coverage

The Narrative Arch

1.      Question / Problem

2.      Logical Sequence / Narrative

3.      Conclusion / Resolution

Two key aspects tie together great story telling – Time and Sequencing

Connected Scatterplot with different segments annotated show an interesting way to show sequencing

-        Napoleon’s march by Minard is one of the most famous connected scatterplots

-        If Russia invaded France the chart might not work as well as it would not read as a western story (which we interpret as Left to Right)

Comics imply the way you should read them (like scroll-telling – scroll to reveal the story)

-        Good example - http://www.bloomberg.com/graphics/2015-whats-warming-the-world/

Animated storytelling like Hans Rosling and demographic change is compelling to the audience – showing grouping / hierarchies is an effective way to communicate the points (ie showing variation within regions)

Zeigarnik effect – theory of information only being retained whilst there is no interruption. Not revealing the answer is a way to make something more memorable – ie a cliff-hanger

-        Memory - http://www.amazon.com/Made-Stick-Ideas-Survive-Others/dp/1400064287 book recommendation from Robert

-        The way to pass on knowledge was by telling stories to pass on memories. Facts / memories have to connect together

-        In academic circles – the computer is the memory in analysis but you need to make someone remember the thing you told them if you want them to make a decision from the information you have shared

Robert Kosara will do a separate talk on how to take Media visualisations and build in Tableau at TC15 in Las Vegas – get it in your diaries now!

Tableau on Tour - Nik Stoychev from easyJet - To the Moon and Back (every day)

1,400 - 1,700 flights per day between 26 bases

Flying A319s and A320s

The planes generate a lot of data: ACARs (event triggers), Fuel Data, WOAR (Quick Access Recorder), Flight Planning Data

The data sources are numerous and varied. The plane often transfers the data when on the ground.

The co-ordinates for the planes used to have to be manually input but this is now automated through ACARs

Nik uses Alteryx to manipulate the data ready for visualisation in Tableau. The analysis done feeds in to the fuel efficiency overheads as EasyJet (as with all other airlines) have to carry at least x% of fuel (number not given) as a safety reserve.

Flight path accuracy used to be quite poor. It was previously enhanced with Excel. Now the work is done in Alteryx to make the process less manual and more efficient.

Bird strikes are shown as a slide of an Angry Bird being fired at an EasyJet plane – best slide so far this conference.

-        The desired goal is to match bird migration patterns with flight direction so the planes don’t get hit (and need to be serviced) and birds don’t die.

Lots of the reporting completed is for the benefit of the regulator as much as the business. Proving flight safety is very important and mapping the data against normal performance under stressed conditions (ie one engine only) what would happen?

Keynote - Dr Ben Goldacre

Keynote 2 – Dr Ben Goldacre – Author of Bad Science and Bad Pharma

1.      Big data (Dr Ben’s term not mine) is used to avoid falling victim to noise

Probability and distribution is vital when working average

Funnel plots are useful to demonstrate this pattern

“Everyone likes to think they are ‘just above’ but only half of us can be above the median”

Systematic Reviews (Meta-Analysis) are the best tests (next is randomised control trials) that you can run all the way through to ‘Ideas, Opinions and Editorials’

-        Science is built on test rather than authority as it should be about clarity and evidence rather than thought

-        PHDs can be bought – Ben did for his dead cat!

Scientific studies get blown out of proportion – findings are often laboratory based and can’t be related to human-world reality

Correlation and Causation – normally at the heart of the issues of bogus claims

Running true Randomised Trials is difficult to close off any other factors

-        “People take a really long time to die and that’s really annoying for medical research”

-        Can’t gloss over the issues of your data set – you have to highlight

Randomised control trial example

-        200 people with headaches

-        Randomly split them in to 2

-        Half get the new pill, half get the old pill

o   Get scores to the changes

Mapping Drug Prescription variance across the UK is very different. In one of the most advanced countries where medical treatment is near at hand, there should be little variation as the decisions should be based on data and not opinion of medical administrators.

The same information about small chances to improved survival rates are inferred differently by different people

-        Chemotherapy usage to increase life expectancy a little or prolong life further for example

Dr Ben argues for providing patients with evidence and information to allow the patient to make logical choices

-        Doctors are given targets that are driving their behaviour that are directly effecting their patients

Should be monitoring the impact of drug assignment on the public by just collecting data about the outcome that just doesn’t get monitored.

Need better data; need better data hygiene to make truly informed decisions and resolve uncertainty

-        There is no urgency to fix this issue as doctors have to learn to ignore the human side

-        The lack of empathy is what is causing the impotence of the decision

Tableau on Tour - Optimising Dashboard Performance - Mrunal Shridhar

You need to start thinking about performance right from the start of your design. If you leave it to the end, it is probably too late.

Basic principles – “it sounds like I’m being a parent. I’m just being practical”

1.      Everything in moderation

2.      If it isn’t fast in the database, it won’t be in Tableau (unless you are using Extracts)

3.      If it isn’t fast in desktop, it won’t be fast in Server

.tde’s up to a few hundred million rows of data – don’t replace your data warehousing solutions

Flat files are opened in a temporary location and therefore doesn’t make anything faster. It’s using RAM. Use an extract to apply indexing.

Server will only beat Desktop when you are hitting the server cache (remember folks, server caching has improved a lot in v9)

4 major processes in desktop:

1.      Connect to data

a.      Native connection vs Generic ODBC (use the driver so it is fast and robust)

                                                    i.     Slow loads could be to a lack of referential integrity

                                                   ii.     Custom SQL is respected by Tableau and avoids join culling etc

2.      Executing Query

a.      Aggregations, Calculated Fields and Filters

b.      Calcs – use Boolean instead of IF? Remove String manipulation and DATEPART()

c.      Filters – often the culprit of slow performance

3.      Computing Layout

a.      Marks, Table Calcs and Sorting

b.      Adding labels and working out if the labels are overlapping that is likely to take a long time

c.      Table Calcs are happening locally so consider pushing back to the data source

4.      Computing Quick Filters

a.      If something isn’t likely to change than having to populate the list of filter options. Dropdowns and wildcard are better as they don’t need to be pre-populated.

Visual Pipeline

Query > Data > Layout > Render

1.      Query – query database, cache results

2.      Data – Local data joins (location data from Tableau joining together with data set), Local calcs, local filters, Totals, Forecasting, Table Calcs, 2^nd Pass filters, Sort

3.      Layout – Layout Views, Compute Legends, Encode marks

4.      Render – Marks, Selection, Highlighting, Labels

Parallel aggregations in v9 really make a difference

External query cache (aka persistent query cache) – the cache is being written to the disk

Multiple data engines – have helped but Query Fusion will assist by working out the common dimensions / aggregations and then working out locally what data is needed for each visualisation

The visual pipeline allows you think about what is happening.

To put the measure on level of detail will help with speed of interactivity

Mrunal uses 144 million rows of flight data to explore performance issues

-        Shows full list for filtering (expensive) and three quick filters (all having to be queried for each stage)

-        Relative date filter or range date filters are faster than date part filters

Using views for filters improves performance and the use of dashboard actions make life faster

Adding parameterised filter to the data source moves it up in the order of operations making your data source smaller, sooner

Mrunal and I will disagree about what the better User Experience is between filters and actions. When labelled well, I personally think dashboard actions make for a lot better experience and keeps you focused on the dashboard rather than the tool.

Aggregate to ‘Visible Dimensions’ is a great data granularity saver. ‘Hide All Unused Fields’ make the data set thinner.

Tableau on Tour - One Shade of Orange - Paul Chapman

Paul is co-host of the London Tableau User Group

BI Journey

EasyJet profit by seat of £8.12

65 million passengers per year, with 85% on time performance. 1,500 staff in HQ

Two key values – Safety first, Customer Focus

BI theme – getting the whole picture for agile decision making

3 years ago – data and reporting outsourced to 3^rd parties (read “slow & expensive”)

Could only get part of the story as the data sets were too large for the data set being analysed

2 desktop licence proof-of-concept

EasyJet partnered with The Information Lab for server deployment, training, mapping support etc

Focused on rolling out to ‘Purple people’ (a mix of analytics and business skills)

CEO asked for the demo to be in their own data – Paul made it so it already was!

Paul wanted to change standards of reporting to create consistency and introduce visualisation best practice

Starting to look at Alteryx to support the Tableau work

Viz Standards

Paul condensed down Stephen Few’s guidelines to create better analytics

Orange, Grey and Blue introduced to get away from Red, Amber and Green

Information Buttons on dashboards to help Consumers make the most of their data

What is going to be different for the CEO? Paul’s honest answer was not a lot apart from speed. Paul highlighted the 4 or 5 chains of command the request goes through to cobble together the ‘beautified question’.

-        “Tangible changes to the way that our analysts work”

Live Demos

Paul took us through a full safety briefing before introducing his live demos

Blending aircraft communcations with PlaneFinder.net API to track routes actually flown

Using Tableau to show the difference between expected journey vs actually flown (timing and fuel usage)

Using routes against maps to see how the pilot is making choices to avoid noise pollution for wealthy areas

Allocated seating – 3 price rows when introduced

Analyse was completed on these numbers. Paul and the team used Tableau to show a custom background image to show where people were sitting on a flight. Different images used to show the different planes.

People are prepared to pay to sit as far forward as possible for the least price.

EasyJet board use their iPads instead of their laptops to consume their dashboards

Paul is testing deploying dashboards through the Apple Watch to keep decision makers close to their data / information.

Birdstrikes are an issue for EasyJet and are therefore monitoring when and where they are happening to insure the correct maintenance is being done to their aircraft.

Showing the impact of strikes by French Aircraft control can help the company understand how to respond to such issues. 

Tableau on Tour London 2015 - Keynote

Opening Keynote

Andy Cotgreave takes the stage…

Attendees arriving from all over the UK, Europe and in to the Middle East and Africa (yes there was a viz to prove that!)

James Eiloart – Extending Our Senses, Unleashing the Human Intellect with Tableau

Making discoveries with data is what making working with data so exciting

Neil DeGrasse-Tyson describes discovery using the metaphors of light – what if we could look through infra-red or night-vision

-        Tolame in Ancient Egypt sees that stars are moving across the night sky and therefore assume everything is rotating round the Earth. He was limited by technology that couldn’t test his theory further.

-        17^th Century – two dutch inventory use a convex lens and concave lens in the same tube. They just invented the Telescope

-        Later in that century – Galileo turns the telescope to the night’s sky. Technology helps the findings and discoveries grow even further

-        1920s – US Prohibition kicks in – Hubble discovers stars are spread across the universe. Hubble had the same data set (the night’s sky) but he had the technology to support his discoveries to see the infinite (discuss!!) expanse

Francois Ajenstat – new in v9 – Smart Meets Fast

Faster performance, smart maps, LoD expressions, data preparation and New Server & Online

The developments keep coming!

Tableau 9.0 adopted 70% faster than 8.2 (the Mac and R release!)

Tableau Online ‘Analytics in the Cloud’

-        SAML support, SSL Connectivity, SSAE-16

-        Live DB Connectivity

-        Online Sync

-        Custom Logos, Embedding

Tableau Online – is Tableau Server but just a hosted version so it gives you online authoring too (desktop in your browser)

For on-premise synchronisation – there is now the ability to sync to the premise rather than having to go to the cloud for each update

Tableau 9.1 updates that are coming

-        Enterprise – 2-way SSL, Sync Active Direcotry on a Schedule, Auto Update (update your desktop)

-        Data – SAP Improvements (SAP HANA Single-Sign-On, Prompts, SAP BW Extracts), Support for Google Cloud SQL & Microsoft Azure DW, Adding a Web Data Connector

-        Analytics – Updates to the Analytic pane

-        Mobile – 9.1 app update will be a big step forward

Web Connector to Google Sheets, Facebook stats etc opens up a whole load of possibilities for new visualisation and analysis. Allows Tableau to link in to Quandl too (Francois showing off Craig’s Quandl connector)

Analytics

-        Median with 95% Confidence Interval (listening to customer feedback)

-        Calculation Editor now in Filter dialogue box

-        Allowing the map to stop paning/zooming

-        Radial map selector – now showing meters / miles on the radial

Mobile

-        Offline sync allows you to explore your data on the move

-        Your favourites will be offline sync’d by manual log-in but auto-updates coming in later versions

-        App is a compliment to the server but the new features will only work with v9.0 or v9.1

The Information Lab live blogging from the Tableau On Tour conference in London